CN110535562A - A kind of filtering type dimensionality reduction interpretation method - Google Patents
A kind of filtering type dimensionality reduction interpretation method Download PDFInfo
- Publication number
- CN110535562A CN110535562A CN201910779678.7A CN201910779678A CN110535562A CN 110535562 A CN110535562 A CN 110535562A CN 201910779678 A CN201910779678 A CN 201910779678A CN 110535562 A CN110535562 A CN 110535562A
- Authority
- CN
- China
- Prior art keywords
- packet
- coding
- decoding
- layer
- degree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
- H03M13/3761—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35 using code combining, i.e. using combining of codeword portions which may have been transmitted separately, e.g. Digital Fountain codes, Raptor codes or Luby Transform [LT] codes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0045—Arrangements at the receiver end
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0057—Block codes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0061—Error detection codes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/0078—Avoidance of errors by organising the transmitted data in a format specifically designed to deal with errors, e.g. location
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0045—Arrangements at the receiver end
- H04L1/0047—Decoding adapted to other signal detection operation
- H04L1/005—Iterative decoding, including iteration between signal detection and decoding operation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L2001/0092—Error control systems characterised by the topology of the transmission link
- H04L2001/0095—Ring
Abstract
The present invention provides a kind of filtering type dimensionality reduction interpretation methods, solve the coding packet of the decoding algorithm dependency degree 1 of RLT, the problem of loss or error code occurs during the coding packet of degree 1 is in dissemination channel, or there are problems that becate in generator matrix, be then easy to cause decoding failure.The present invention carries out mathematical analysis to becate problem, and defined analysis " stringent becate " problem, is further used to stringent becate on this basis, proposes FDR decoding algorithm, makes up the decoding defect of RLT, improves decoding success rate.
Description
Technical field
The present invention relates to network communication technology fields, relate in particular to a kind of filtering type dimensionality reduction interpretation method, are used for water
The transmission of lower sensor network multimedia information.
Background technique
Underwater sensor network (UWSN) in environmental monitoring, resource exploration, disaster prevention etc. due to having a wide range of applications
Prospect causes the very big concern of researcher.Due to the difference of application, people for underwater data format demand not only
It is confined to the text information of lightweight.Size is that picture even underwater real-time video of up to several M of several K etc. is various forms of
The transmission of multimedia messages is increasingly becoming the research interest of scientific worker.The the data volume of information the big often to transmit data
Real-time and reliability requirement it is stringenter, however submarine link is because there are more unfavorable factors to make its less reliable.
Therefore, the Transmitting of Multi-media Information of underwater data and reliable transmission technology are closely bound up.Digital fountain code (digital
Foundation codes, DFC) it is a kind of efficiently without code rate Error Correction of Coding, the coding mode fixed compared to code rate has
More reliable excellent properties, the underwater acoustic network severe for communication environment show good adaptability.
There are problems for the prior art: the decoding algorithm of digital fountain code mainly includes two kinds: BP algorithm and GE algorithm.BP
(belief propagation, belief propagation) algorithm is excessively high to the coding packet dependence of degree, so that the utilization rate of encoded information
It is lower.The decoding process of GE (Gaussian elimination, Gaussian elimination) algorithm calculates multiple relative to solution linear equation
Miscellaneous degree is higher.
Applicant proposed a kind of digital fountain code-RLT encoding and decoding algorithm being simple and efficient, RLT is encoded using recurrence, is compiled
Decoding efficiency is high, and successfully decoded probability also greatly increases.But the decoding algorithm of RLT wants the coding packet of dependency degree 1 eventually, if
Loss or error code occur during dissemination channel for 1 coding packet of degree, or there are problems that becate in generator matrix, then are easy to lead
Cause decoding failure.The present invention carries out mathematical analysis to becate problem, on this basis defined analysis " stringent becate " problem, into
One step is used to stringent becate, proposes FDR decoding algorithm, makes up the decoding defect of RLT, improves decoding success rate.
Summary of the invention
The invention solves key technical problem be that the decoding algorithm of RLT wants the coding packet of dependency degree 1 eventually, if
Loss or error code occur during dissemination channel for 1 coding packet of degree, or there are problems that becate in generator matrix, then are easy to lead
The problem of causing decoding failure.
In order to solve the above technical problems, the present invention adopts the following technical scheme: a kind of filtering type dimensionality reduction interpretation method, special
Sign is: including the following steps:
(1) decoding process can be started when receiving first coding packet using the receiving end of FDR decoding algorithm, as long as
Between its corresponding original packet set for participating in coding original packet set for participating in coding corresponding with the coding packet being successfully received
There are true inclusion relations, it is not necessary to which the coding packet of degree of waiting 1, this shortens decoding time to a certain extent；
(2) to improve the successfully decoded probability of FDR, increase big angle value under the premise of reasonable quantity of the degree for 1 coding packet
Packet probability is encoded, so that the average degree of coding packet is smaller, combines undesirable covering problem, it is more in original packet quantity
In the case where not only can control decoding and be unlikely to too complicated, but also can guarantee higher successfully decoded probability；
The degree distribution of the coding packet uses following design:
Wherein, k indicates that each data block includes k raw data packets, and encoder encodes k raw data packets, d
It is the degree of coding packet, d ∈ { 1,2 ..., k }；Ω (d) degree of a representation is the probability distribution of d；
(3) FDR decoder uses hierarchical design thought, and the angle value for encoding packet has 5 kinds, is respectively as follows: d=1, d=2, d=3,
D=4, d=k, decoder are designed as 5 layers: l therewith1,l2,l3,l4,lk, the coding packet of corresponding angle value, the volume in layer are stored respectively
Code packet both includes the coding packet received, further includes the secondary or multiple decoding packet generated after exclusive or is handled, it should be noted that
: lkLayer storage angle value range be d ∈ (4, k] coding packet or decoding packet.
Above-mentioned FDR interpretation method, using following decoding process:
(1) FDR algorithm regulation transmitting terminal first sends the coding packet Y of d=k for after the completion of data encodingn, then successively send
The coding packet of d=4, d=3, d=2, d=1, the decoding process of receiving end are；
(2) first coding packet Y is receivedn, because of d (Yn)=k, puts it into lkLayer；
(3) the coding packet Y that will be successfully receivediFrom lkLayer starts, and compares one by one with the coding packet in layer, judges whether to meet
Exclusive or condition；
(4) if YiPacket Y is encoded with certainjMeet exclusive or condition and d (Yi)<d(Yj) thenThen by angle value compared with
Small YiIt is put into its respective layer；The biggish Y of angle valuejIt directly abandons, no longer saves；Calculate the angle value d of the secondary coding packet generated
(Ysec), put it into corresponding layer；If the coding packet is until l1Layer does not meet exclusive or condition always, is put according to angle value
Enter respective layer；
(5) 3-4 step is repeated, until all codings packet receives, if l at this time1The vector S of layer includes all original packets,
It is as successfully decoded；
(6) if still thering is original packet not restore, by l1The original packet of layer is successively and l2,l3,l4The coding packet of layer carries out exclusive or,
Until successfully decoded.
Detailed description of the invention
Fig. 1 is off two kinds of situations of collection；
Fig. 2 is that the decoding of LT code terminates phenomenon；Wherein, white circle is raw data packets, and black circle is coding packet；
Fig. 3 is 4 yuan of becate examples；
Fig. 4 is FDR decoder concept map；
Fig. 5 is formula 1-6；
Fig. 6 is the degree distribution function operation of optimization；
Fig. 7 is that encoding and decoding complexity compares.
Specific implementation method
Application method and device in the following embodiments of the invention patent are unless otherwise specified conventional method and device；
Equipment used, reagent are the conventional equipment and reagent of Reagent Company's purchase.For purpose, the technical solution for making the invention patent
It is clearer with advantage, the specific embodiment of the invention patent is described in detail combined with specific embodiments below.These
The example of preferred embodiment is illustrated in a particular embodiment.
Here, it should also be noted that, in order to avoid having obscured the technical side of the invention patent because of unnecessary details
Case illustrate only the technical solution and/or processing closely related with the scheme of patent according to the present invention in embodiment and walk
Suddenly, the little other details of relationship are omitted.
Embodiment 1
The recurrence RLT encoding-decoding efficiency proposed in the past is high.But the decoding algorithm of RLT wants the coding packet of dependency degree 1 eventually, such as
Loss or error code occur during dissemination channel for the coding packet of fruit degree 1, then decoding will fail.When decoding terminates in advance, institute
There is still useful coding packet (degree is greater than 1 coding packet) in the coding packet received, the set that these coding packets are constituted is to stop
Only collect.It is explained with two kinds of phenomenons and stops collection problem: 1) line, such as Fig. 1 is not present between certain original packets and any coding packet
(a) shown in, original symbol S1, S2Can be in the hope of by XOR operation, but S3It is not engaged in the coding of any coding packet, so
S anyway3It can not obtain；2) the remaining coding packet angle value not solved is all larger than 1 during decoding, that is, not
The coding packet of presence 1 causes decoding process to terminate.As shown in Fig. 1 (b), remaining coding wraps Y1、Y2、Y3Degree be all larger than
1, it can not continue to decode at this time.
The first situation is as randomly selecting in cataloged procedure caused by original packet carries out exclusive or, because randomly selecting
Method make certain original packets there is certain not selected probability thus directly result in these original packets fail to participate in it is any
The coding of packet is encoded, such case is also referred to as undesirable covering (imperfect coverage, IC) problem；Second situation
It is then the becate problem next to be analyzed.
Becate problem
It defines 1 in generator matrix, if there is such two column, their two rows (or more than two rows) in same position are equal
For " 1 ", these rows being made of " 1 " form the ring of a closure, referred to as " becate ".
If meeting the row that becate defines has two rows, the becate that this two row is constituted is 4 member rings, if such row has
Three rows, then the becate that this three row is constituted is 6 member rings, and so on: assuming that meeting the row that becate defines has k'(2≤k' < k)
A, the becate that they are constituted is (2k') member ring.It is explained for occurring terminating phenomenon during decoding shown in Fig. 2 " short
Ring ".
Y1=S2,It is decoded according to LT code
Process, first degree of finding 1 coding packet Y1, S can be directly translated at this time2And by Y1And S2Between line remove；In next step, by institute
Have and S2Connected coding packet { Y2,Y3,Y4Respectively with S2Make XOR operation, while updating Y2,Y3,Y4Value be XOR operation knot
Fruit, and the line between them is deleted.At this point, remaining coding wraps Y2,Y3,Y4Angle value d be respectively d (Y2)=2, d
(Y3)=2, d (Y4)=3.According to LT code decoding rule, the coding packet of degree of finding out 1 is needed, but the angle value of three coding packets is equal
Meet and be more than or equal to 2, so decoding is forced to terminate.At this point, Y2,Y3,Y4Corresponding generator matrix is as shown in Figure 3.
Include the becate of 2 rings a length of 4 in the generator matrix: Y3, Y4 corresponding two is listed in the first row and the third line is all
" 1 ", so the first row and the third line constitute 4 yuan of becates；Similarly, Y2,Y4Corresponding two are listed in the third line and fourth line
It is " 1 ", so this two row constitutes second 4 yuan of becate.By taking 4 member rings as an example, it is assumed that the coding parameter of LT code be (n, k,
Ω), k is original packet number, and n is coding packet number, and Ω is the degree distribution function of coding packet, in the generator matrix G of n × k rank
The generating probability for spending i symbol is Ωi, the generating probability of degree j symbol is Ωj；So, angle value is respectively i in matrix G, j (i > 1,
J > 1) certain two column constitute 4 yuan of becates probability such as 5 formula of attached drawing (1) shown in.
Stringent becate
Present the application provides stringent becate definition: defining 2, in generator matrix, if there is such two column, they are in phase
It is " 1 " with two rows (or more than two rows) on position, and wherein row of the column in addition to this two rows (or multirow) remaining position
On be " 0 " entirely, these be made of " 1 " row is formed one be closed ring, we term it " stringent becates ".Assuming that meeting tight
The row that lattice becate defines has k'(2≤k' < k) it is a, the becate that they are constituted is (2k') member ring.So, angle value is j in matrix G
The column that the column of (j > 2) and a certain angle value are 2 are constituted shown in probability such as 5 formula of attached drawing (2) of stringent 4 yuan of becates.So, matrix G
The column and a certain angle value that middle angle value is j are that the column of m (m < j) constitute probability such as 5 formula of attached drawing (4-3) institute of stringent 2m member becate
Show.
Embodiment 2
The main decoding algorithm of fountain codes is there are two types of type: belief propagation algorithm (belief propagation, BP) and
Gaussian elimination algorithm (Gaussian elimination, GE).The coding of BP algorithm degree of needs 1 starts as iteration, is distributed to degree
Designing reasonable code word then has brilliant performance, but the coding for being 1 due to excessively relying on degree, the utilization rate of encoded information
It is not high, and centering long code word calculating speed is slower；GE algorithm uses maximum-likelihood decoding algorithm (maxi-mum to liner code
Likelihood, ML), it is equivalent to solution linear equation, various fountain codes decodings is can be applied to, can make full use of encoded information,
But since its computation complexity is higher, with the increase of input message length, efficiency will decline to a great extent.Two kinds of decoding algorithms are each
There is its scope of application, decoding algorithm need to be reasonably selected according to actual needs, also it is contemplated that combining two kinds of algorithms simultaneously improves decoding
Efficiency.
The present invention proposes a kind of filtering type dimensionality reduction decoding algorithm --- FDR algorithm (filtering dimension
reduction decoding algorithm).FDR algorithm wraps it using the coding that can constitute stringent becate in generator matrix
Between xor operation, packet that generation degree is 1 or degree higher coding packet be subjected to " dimensionality reduction " dropped even if handling its degree
It is low.Will not be as BP decoding algorithm, FDR decoding algorithm, which not only relieves, encodes the degree 1 generated from transmitting terminal encoder
The dependence of packet, can be with the degree of fast reducing height number encoder packet to further decrease decoding complexity.
Decode thought
In the analysis of upper section becate problem, when there is Fig. 2 LT code decoding termination phenomenon, three codings wrap Y2,Y3,Y4
24 yuan of stringent becates are contained in the generator matrix of composition.Next it is carried out from the angle for the original packet set for participating in coding
Analysis.The collection of the corresponding original packet for participating in coding of coding packet Y2 is combined into { S3,S4, Y3The corresponding original packet for participating in coding
Collection is combined into { S1,S3, Y4The collection for participating in the original packet of coding is combined into { S1,S3,S4}.These three set exist significantly comprising closing
System: If by Y2And Y4Carrying out XOR operation can be obtainedBy Y3And Y4It is available to carry out XOR operationBy S1And Y3It is available to carry out exclusive orOriginally using BP decoding algorithm can only dependency degree 1 Y1Translate S2, S2Decoding operation amount CS2=1, and S1,
S3,S4It can not solve at all, it is believed that obtain S1,S3,S4Decoding operation amount CS1, CS3, CS4Close to infinity: CS1,
CS3,CS4→ ∞, the decoding operation amount for then obtaining all original packets is Call=CS1+CS2+CS3+CS4→∞.But pass through volume
Exclusive or between code packet, can solve three original packet S1,S3,S4。S1,S2,S3,S4Decoding operation amount be respectively C'S1=1,
C'S2=1, C'S3=2, C'S4=1, then the decoding operation amount that all original packets are successfully decoded is C'all=C'S1+C'S2+C'S3
+C'S4=5=Call。
It is seen on surface, waste when stringent becate is a kind of coding, the decoding process of packet is encoded simultaneously to traditional dependency degree 1
It is valueless, but decoded mode is made certain change, stringent becate is used, the contribution of stringent becate will be very important,
Possibly even become the committed step of the remaining surplus failed original packet of decoding.Therefore, have to draw a conclusion:
As long as the two column degrees for constituting becate in 1 generator matrix of conclusion are different, and that lesser column of angle value are in addition to structure
At the row of becate there are 1, other rows can carry out XOR operation between the coding packet corresponding to them there is no when 1.The two
The angle value for the new coding packet (referred to as secondary coding packet) that exclusive or generates is necessarily smaller than alternative one, or even all than the two angle value
It is small.If the two angle value difference 1, their exclusive or directly obtain the packet that a degree is 1；If the two degree differs by more than 1, warp
It crosses after exclusive or, the angle value of that biggish column of degree can be reduced to the difference of the two degree.
LT code decoding process will carry out after receiving certain amount coding packet；RLT code makes improvements herein, data encoding
After the coding of degree of transmission 1 packet first, the decoding process of RLT code starts at once when receiving coding and wrapping, but both are translated
Code mode is all the coding packet starting decoding of dependency degree 1 eventually.And the decoder of FDR algorithm is used, receiving first coding packet
It can start to decode after (no matter whether angle value is 1).If first coding received using the receiving end of FDR decoding algorithm
Packet is Y4, the second coding packet received is Y3When, the original packet set of both receiving end comparison participates in coding, two set are deposited
In true inclusion relation, receiving end can start decoding process, carry out XOR operation to two coding packets.
The present invention proposes a kind of filtering type dimensionality reduction interpretation method (filtering dimension reduction
decoding algorithm, FDR)。
Embodiment 3
Next the design and decoding process of decoder are mainly introduced.Several parameters in FDR decoding algorithm given first:
Define 4k input symbolic vector S:S={ S1,S2,...,Sk, n of coding generation are carried out to k input symbol
Coded symbol vectors Y:Y={ Y1,Y2,...,Yi...,Yn, coding therein wraps YiAngle value be expressed as d (Yi)；FDR algorithm
Coding packet is divided into two types, in addition to the n coding generated by k original packet through transmitting terminal encoder wraps Y1,Y2,...,Yn
Except, it further include the secondary coding packet that coding packet generates in the decoder of receiving end with other coding packet exclusive or, in order to distinguish,
Secondary coding packet YsecIt indicates, then the angle value of secondary coding packet is expressed as d (Ysec)；One coding packet (no matter Y1,Y2,...,
YnIn some YiOr some secondary coding packet Ysec) the ID collection of the corresponding original packet for participating in its coding is combined into T.
The design of decoder
Decoder uses hierarchical design thought according to the angle value range of coding packet.In degree distribution function proposed by the present invention
In, the angle value d (d ∈ Z) for encoding packet has 5 kinds, is respectively as follows: d=1, d=2, d=3, d=4, d=k.Decoder designs therewith
It is 5 layers: l1,l2,l3,l4,lk, as shown in Fig. 4 decoder concept map.What every layer of decoder was stored respectively is the volume of corresponding angle value
Code packet, it further includes that the coding packet from receiving end enters that herein, the coding packet in layer, which had both included the coding packet from receiving end,
The secondary coding that already present (being introduced into decoder) coding packet exclusive or generates when flowing between the layers after decoding and in layer
Packet.For example, l2The coding packet that all degree are 2 is housed, there may be the degree from transmitting terminal received in these coding packets
2 coding packet, it is also possible to which the coding packet of the coding packet and a degree 1 that have a degree 3 is in l1Layer exclusive or generate degree 2 it is secondary
Coding packet.Furthermore it should be noted that here, lkLayer storage angle value range be d ∈ (4, k] coding packet, spending can be with for the coding packet of k
Packet, which is encoded, with any one carries out exclusive or, then the coding packet of degree k (assuming that k > 8) and degree t (t=1,2,3,4)
The angle value for the secondary coding packet that packet exclusive or generates is encoded as k-t, and k-t > 4, therefore angle value is greater than 4 secondary codings for being less than k
Packet is put into lkLayer.With the coding packet of angle value d=k=10 in l4For XOR operation occurs for layer, the secondary coding packet Y of generationsec's
Angle value is d=k-4=6, so by secondary coding packet YsecIt is placed on lkLayer.Existence form of the coding packet in decoder is a pair of
Key-value, key --- ID set T, value that the coding wraps the corresponding original packet for participating in constituting its coding --- volume
Code packet.For example, coding packetS1,S2,S3Corresponding ID is respectively 0,1,2, TYi={ 0,1,2 }, then Yi
In l3The existence form of layer is { 0,1,2 }-Yi.It after coding packet enters decoder, flows downward, successively passes through each layer of decoder,
When reaching a certain layer, if there are true inclusion relations by the T that some coding wraps in the T and current layer of the coding packet, in other words
It says, all raw data packets spent in small coding packet have been also involved in the XOR operation that the biggish coding of degree wraps, and (this is exclusive or
Condition), then the biggish coding coating of angle value is updated to secondary coding packet, and angle value is lowered after the two XOR operation.
This process be equivalent to the coding packet big to angle value be filtered, " dimensionality reduction ", reduce height number encoder packet angle value, in certain journey
Decoding progress is accelerated on degree.Angle value is larger or the coding packet of angle value very little is transported in certain layer and other codings raw exclusive or of contracting for fixed output quotas
The probability of calculation is bigger.
Theoretically, the l of encoder2,l3,l4,lkLayer, with below it remove l1There is no true for each T set in all layers
Inclusion relation.Since FDR decoder have been at decoded state, until l receiving first coding packet1Layer comprising entirely to
S is measured, is considered as successfully decoded.
FDR decodes process
FDR algorithm provides that transmitting terminal by after the completion of data encoding, first sends the coding packet Y of d=kn, then successively send d=
The coding packet of 4, d=3, d=2, d=1.
(1) first coding packet Y is receivedn, because of d (Yn)=k, puts it into lkLayer；
(2) the coding packet Y that will be successfully receivediFrom lkLayer starts, and compares one by one with the coding packet in layer, judges whether to meet
Exclusive or condition；
(3) if YiPacket Y is encoded with certainjMeet exclusive or condition and d (Yi)<d(Yj) thenThen by angle value compared with
Small YiIt is put into its respective layer；The biggish Y of angle valuejIt directly abandons, no longer saves；Calculate the angle value d of the secondary coding packet generated
(Ysec), put it into corresponding layer；If the coding packet is until l1Layer does not meet exclusive or condition always, is put according to angle value
Enter respective layer；
(4) (2)-(3) step is repeated, until all codings packet receives, if l at this time1The vector S of layer includes all original
Packet, it is as successfully decoded.
(5) if still thering is original packet not restore, by l1The original packet of layer is successively and l2,l3,l4The coding packet of layer carries out exclusive or,
Until successfully decoded.
Optimization degree distribution function
The degree distribution function of digital fountain code influences encoder complexity, is related to code efficiency.Reasonable degree distribution is answered
The selection probability for increasing big angle value coding groups under the premise of the reasonable quantity in 1 coding groups of guarantee degree, makes in this way by spending
The coding groups degree that be averaged that distribution generates is smaller, combines undesirable covering problem, guarantee is to the good of all original packets
Good covering.
Give k raw data packets, following 5 formula of attached drawing (4) of degree distribution design: wherein in formula (4) ρ (d) expression
Formula is 5 formula of attached drawing (5): the expression formula of τ (d) is 5 formula of attached drawing (6) in formula (5).
When d=1, the quantity for encoding packet is related to k, and original packet quantity k is bigger, and 1 coding packet quantity of degree is more, is more conducive to
Accelerate decoding speed；When d=k, coding packet quantity is 1, guarantees that all original packets are involved in the coding of the packet, avoids
Undesirable covering problem.The design of d=2, d=3, d=4 are general by degree 2, degree 3, the selection of degree 4 from ideal solitary wave distribution
Rate is set smaller than 1 number, so that the optimization of degree distribution to be transformed into the design problem of constant.
6 operation of attached drawing is known by the degree distribution function optimized:
In general,It is a number no more than 0.3, so λ ≈ 3.57, the decoding complexity of improved RLT code are
3.57, it is unrelated with original packet quantity k.The decoding complexity of RLT code, improved RLT code and the fountain codes using the distribution of tradition degree
Compare as shown in 7 table 1 of attached drawing.
The utility model has the advantages that the application is a kind of filtering type dimensionality reduction interpretation method, have the advantage that
Receiving end can start decoding process when receiving first coding packet, as long as its corresponding original for participating in coding
There are true inclusion relations between the packet set that the begins original packet set for participating in coding corresponding with the coding packet being successfully received, just not
Must degree of waiting 1 coding packet, this shortens decoding time to a certain extent.FDR algorithm is tight using that can constitute in generator matrix
Xor operation between the coding packet of lattice becate, the packet or carry out the higher coding packet of degree at " dimensionality reduction " that generation degree is 1
Even if managing its degree reduces.Will not be as BP decoding algorithm, FDR decoding algorithm is not only relieved to be encoded to from transmitting terminal
The dependence for the coding packet of degree 1 that device generates, can be multiple to further decrease decoding with the degree of fast reducing height number encoder packet
Miscellaneous degree.
Simulation comparison is carried out with RLT code in terms of successfully decoded probability using NS3 emulation platform, and has been counted using FDR
Algorithm generates secondary decoded number in 100 times or so data transmission.A large amount of simulation results show, FDR algorithm is compared with RLT
Algorithm performance is more excellent.Effect of the present invention is ideal, and application prospect is good.
Embodiment 4
The present embodiment provides a kind of underwater sensor network Transmitting of Multi-media Information methods, include the following steps:
After the degree distribution for solving the problems, such as digital fountain code and coding and decoding, next need to realize based on coding and decoding scheme
The reliable transmission controlling mechanism of real time communication between node.Underwater sound Modem is usually operated at semiduplex mode, sends and receivees
Be unable to simultaneously perform, one it is available transmission controlling mechanism should be able to avoid in transmission state node-node transmission packet and
Caused transmission-reception interference.Most access-control protocols so far use RTS/CTS handshake mechanism dynamic coordinate section
Point transmission.The rate that underwater sensor network generates data is 1-5bps, and the packet length of optimization is 100 bytes, and RTS/CTS
Wrap a length of tens bytes.Therefore, compared with data packet, RTS/CTS frame length be not it is very short, using RTS/CTS handshake mechanism
Bring income is not significant.On the contrary, the features such as considering underwater acoustic channel narrow bandwidth, long delay, RTS/CTS, which shakes hands, to be reduced
Channel utilization, network throughput, extend end-to-end delay.Therefore, based on the reliable transmission controlling party of our encoding and decoding
The transmission mechanism that no RTS/CTS shakes hands can be used in case.
Raw data packets are divided into the block of k size by source node first, i.e., every data block includes k raw data packets.Source
Node encodes k original packet.Sensor network under water sends the data block comprising 50 optimization packets and needs
Time be about 60 seconds, can satisfy the transmission time being limited between two nodes.By suitably setting block size control transmission
Time wraps so that receiving node be enable to receive enough codings to reconstruct raw data packets, with realizing block-by-block, hop-by-hop reliable
Transmission.
In the block-by-block hop-by-hop transmission based on encoding and decoding, those nodes for sending packet considered to be in transmission state.
In order to avoid the synchronization conflicts between data frame-ACK, over-redundancy is reduced, is rolled between efficiency of transmission and fairness
Inner feelings, we define two transmission constraint conditions: 1) the maximum data packet quantity for allowing to send in the primary transmission stage is N.2) same
The minimum interval in one node two transmission stage is Ta, wait TaExpired node, which is considered being in transmission, avoids rank
Section.In view of the transfer delay that acoustics Modem sends and receives state is larger, usually the second grade.Here settable Ta=2RTT
Instead of Ta=RTT.
In first transmission stage of a data block, sending node sends N number of coding packet.Sending node is switched to later
Reception state waits the feedback information of receiving node to be received.
Feedback information from receiving node includes the quantity N of the frame receivedrAnd the original packet number k that cannot be reconstructed1。
K in original degree distribution is replaced with k in second transmission stage of the data block by sending node1, and to k1A original packet
It carries out second to encode, generates and send N1A coding packet, switching state waits feedback information later.Repeatedly, Zhi Daoyi
Until code success.
The above is only the specific embodiment of the application, it is noted that for the ordinary skill people of the art
For member, under the premise of not departing from the application principle, several improvements and modifications can also be made, these improvements and modifications
It should be regarded as the protection scope of the application.
Claims (2)
1. a kind of filtering type dimensionality reduction interpretation method, characterized by the following steps:
(1) decoding process can be started when receiving first coding packet using the receiving end of FDR decoding algorithm, as long as its is right
Exist between the original packet set for the participation coding the answered original packet set for participating in coding corresponding with the coding packet being successfully received
True inclusion relation, it is not necessary to which the coding packet of degree of waiting 1, this shortens decoding time to a certain extent；
(2) to improve the successfully decoded probability of FDR, increase big angle value coding under the premise of reasonable quantity of the degree for 1 coding packet
Packet probability combines undesirable covering problem, in a fairly large number of situation of original packet so that the average degree of coding packet is smaller
Under not only can control decoding and be unlikely to too complicated, but also can guarantee higher successfully decoded probability；
The degree distribution of the coding packet uses following design:
Wherein, k indicates that each data block includes k raw data packets, and encoder encodes k raw data packets, and d is to compile
The degree of code packet, d ∈ { 1,2 ..., k }；Ω (d) degree of a representation is the probability distribution of d；
(3) FDR decoder uses hierarchical design thought, and the angle value for encoding packet has 5 kinds, is respectively as follows: d=1, d=2, d=3, d=
4, d=k, decoder is designed as 5 layers: l therewith1,l2,l3,l4,lk, the coding packet of corresponding angle value, the coding packet in layer are stored respectively
Both include the coding packet received, further include the secondary or multiple decoding packet generated after exclusive or is handled, it should be noted that:
lkLayer storage angle value range be d ∈ (4, k] coding packet or decoding packet.
2. a kind of filtering type dimensionality reduction interpretation method according to claim 1, it is characterised in that: the FDR interpretation method is adopted
With following decoding process:
(1) FDR algorithm regulation transmitting terminal first sends the coding packet Y of d=k for after the completion of data encodingn, then d=4 is successively sent, d
The coding packet of=3, d=2, d=1, the decoding process of receiving end are；
(2) first coding packet Y is receivedn, because of d (Yn)=k, puts it into lkLayer；
(3) the coding packet Y that will be successfully receivediFrom lkLayer starts, and compares one by one with the coding packet in layer, judges whether to meet exclusive or
Condition；
(4) if YiPacket Y is encoded with certainjMeet exclusive or condition and d (Yi)<d(Yj) thenThen by the lesser Y of angle valuei
It is put into its respective layer；The biggish Y of angle valuejIt directly abandons, no longer saves；Calculate the angle value d (Y of the secondary coding packet generatedsec),
Put it into corresponding layer；If the coding packet is until l1Layer does not meet exclusive or condition always, puts it into correspondence according to angle value
Layer；
(5) 3-4 step is repeated, until all codings packet receives, if l at this time1The vector S of layer includes all original packets, is as translated
Code success；
(6) if still thering is original packet not restore, by l1The original packet of layer is successively and l2,l3,l4The coding packet of layer carries out exclusive or, until translating
Code success.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910779678.7A CN110535562B (en) | 2019-08-22 | 2019-08-22 | Filtering type dimension-reduction decoding method |
| AU2020101634A AU2020101634A4 (en) | 2019-08-22 | 2020-08-03 | Filtering dimension reduction decoding method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910779678.7A CN110535562B (en) | 2019-08-22 | 2019-08-22 | Filtering type dimension-reduction decoding method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110535562A true CN110535562A (en) | 2019-12-03 |
| CN110535562B CN110535562B (en) | 2020-08-04 |
Family
ID=68664033
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910779678.7A Active CN110535562B (en) | 2019-08-22 | 2019-08-22 | Filtering type dimension-reduction decoding method |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110535562B (en) |
| AU (1) | AU2020101634A4 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112994847A (en) * | 2021-03-12 | 2021-06-18 | 北京理工大学 | Limited feedback method of online fountain codes |
| US20230055353A1 (en) * | 2022-04-26 | 2023-02-23 | Qinghai Normal University | Data encoding and decoding method for underwater acoustic networks (uans) based on improved online fountain code |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070260957A1 (en) * | 2006-05-03 | 2007-11-08 | Emina Soljanin | Encoded transmission |
| CN101562498A (en) * | 2008-04-18 | 2009-10-21 | 中国移动通信集团公司 | Encoding and decoding transmission method of multi-join data stream cracking as well as device and system thereof |
| CN101902796A (en) * | 2009-05-27 | 2010-12-01 | 华为技术有限公司 | Acquisition degree distributing method, data transmission method and equipment as well as relay system |
| CN102237966A (en) * | 2011-07-11 | 2011-11-09 | 哈尔滨工业大学深圳研究生院 | Digital fountain code decoding method based on degree 2 and high-degree encoding packets |
| CN102255690A (en) * | 2011-07-11 | 2011-11-23 | 哈尔滨工业大学深圳研究生院 | Method for decoding fountain codes based on 2 and 3 degrees of coding packets |
| CN102638332A (en) * | 2012-03-22 | 2012-08-15 | 西安电子科技大学 | Improved LT (language translation) decoding method |
| CN104539387A (en) * | 2014-11-28 | 2015-04-22 | 杜秀娟 | Hop-by-hop reliable transmission control method for underwater wireless sensor network |
-
2019
- 2019-08-22 CN CN201910779678.7A patent/CN110535562B/en active Active
-
2020
- 2020-08-03 AU AU2020101634A patent/AU2020101634A4/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070260957A1 (en) * | 2006-05-03 | 2007-11-08 | Emina Soljanin | Encoded transmission |
| CN101562498A (en) * | 2008-04-18 | 2009-10-21 | 中国移动通信集团公司 | Encoding and decoding transmission method of multi-join data stream cracking as well as device and system thereof |
| CN101902796A (en) * | 2009-05-27 | 2010-12-01 | 华为技术有限公司 | Acquisition degree distributing method, data transmission method and equipment as well as relay system |
| CN102237966A (en) * | 2011-07-11 | 2011-11-09 | 哈尔滨工业大学深圳研究生院 | Digital fountain code decoding method based on degree 2 and high-degree encoding packets |
| CN102255690A (en) * | 2011-07-11 | 2011-11-23 | 哈尔滨工业大学深圳研究生院 | Method for decoding fountain codes based on 2 and 3 degrees of coding packets |
| CN102638332A (en) * | 2012-03-22 | 2012-08-15 | 西安电子科技大学 | Improved LT (language translation) decoding method |
| CN104539387A (en) * | 2014-11-28 | 2015-04-22 | 杜秀娟 | Hop-by-hop reliable transmission control method for underwater wireless sensor network |
Non-Patent Citations (1)
| Title |
|---|
| 柳秀秀 等: "《基于优化Ｒaptor 码的水下传感器网络可靠传输机制》", 《重庆邮电大学学报(自然科学版)》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112994847A (en) * | 2021-03-12 | 2021-06-18 | 北京理工大学 | Limited feedback method of online fountain codes |
| CN112994847B (en) * | 2021-03-12 | 2022-02-11 | 北京理工大学 | Limited feedback method of online fountain codes |
| US20230055353A1 (en) * | 2022-04-26 | 2023-02-23 | Qinghai Normal University | Data encoding and decoding method for underwater acoustic networks (uans) based on improved online fountain code |
| US11722245B2 (en) * | 2022-04-26 | 2023-08-08 | Qinghai Normal University | Data encoding and decoding method for underwater acoustic networks (UANs) based on improved online fountain code |
Also Published As
| Publication number | Publication date |
|---|---|
| AU2020101634A4 (en) | 2020-09-10 |
| CN110535562B (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103348739B (en) | For the method and apparatus supporting Modulation and Coding Scheme to arrange in high throughput wireless system | |
| Koetter et al. | A theory of network equivalence—Part I: Point-to-point channels | |
| CN106713956B (en) | Code rate control and version selection method and system for dynamic self-adaptive video streaming media | |
| Frey et al. | A gamma-based framework for modeling variable-rate MPEG video sources: The GOP GBAR model | |
| CN109347604B (en) | Multi-hop network communication method and system based on batched sparse codes | |
| CN101636983A (en) | Reducing effects of packet loss in video transmissions | |
| CN110535562A (en) | A kind of filtering type dimensionality reduction interpretation method | |
| US9762957B2 (en) | Network coding for multi-resolution multicast | |
| CN106851733A (en) | A kind of adaptive H TTP message compression methods for mobile network's application | |
| Yue et al. | Communication-efficient federated learning via predictive coding | |
| Itahara et al. | Communication-oriented model fine-tuning for packet-loss resilient distributed inference under highly lossy IoT networks | |
| CN109274462B (en) | Image transmission method based on improved online fountain codes | |
| Wu et al. | Transformer-aided wireless image transmission with channel feedback | |
| CN111447044B (en) | Distributed storage method and transmission decoding method | |
| US10097602B2 (en) | Channel-aware caching-aided coded multicast | |
| Keshtkarjahromi et al. | Content-aware network coding over device-to-device networks | |
| US10142049B2 (en) | Near optimal forward error correction system and method | |
| Nguyen et al. | Network coding-based wireless media transmission using POMDP | |
| CN104821881B (en) | A kind of thing network sensing layer data safety fusion method based on compression and Error Correction of Coding | |
| CN107615810A (en) | Header compression system and method for online network code | |
| CN113507467B (en) | Block chain-based hidden information transmission system and method | |
| Rao et al. | CNN-SC decoder for polar codes under correlated noise channels | |
| CN105141386A (en) | Rateless code transmission method suitable for dying channel | |
| CN103427951A (en) | Method for forwarding data with encoding redundancy control | |
| CN115811742B (en) | High-energy-efficiency transmission method for improving reliability of CoAP protocol unacknowledged mode |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |